Drug package inserts are technical and scientific documents that usually do not contain illustrations and are full of technical jargon and terms, small print, and long texts that hinder understanding by patients and correct use of the medicines This project is thus intended to use large language models (LLM) trained with the wording from Brazilian package inserts and through natural language processing (NLP) techniques for selection and extraction of the texts, aimed at creating an alternative format with accessible, easy-to-understand language. It is expected that the trained LLM tool can generate summarized and up-to-date information on medicines, aimed at improving orientation for users. This approach includes a promising perspective since there are few digital resources for this purpose. The end users will be patients in the Unified Health System (SUS) as well as healthcare professionals involved in the drugs’ prescription, dispensing, and administration. Since the proposal also involves creating a dataset, this product can be relevant for data science professionals that work in the public administration and for other research projects that involve specific communities, such as elderly people and people in situations of vulnerability.